gh-139772: Add PyDict_NewPresized() function

Author: vstinner

Doc/c-api/dict.rst

@@ -36,6 +36,18 @@ Dictionary Objects
    Return a new empty dictionary, or ``NULL`` on failure.
 
 
+.. c:function:: PyObject* PyDict_NewPresized(Py_ssize_t size, int unicode_keys)
+
+   Return a new empty dictionary with preallocated items, or ``NULL`` on
+   failure.
+
+   *size* is a hint to preallocate items.
+
+   If *unicode_keys* is non-zero, optimize lookup for Unicode keys.
+
+   .. versionadded:: next
+
+
 .. c:function:: PyObject* PyDictProxy_New(PyObject *mapping)
 
    Return a :class:`types.MappingProxyType` object for a mapping which

Doc/whatsnew/3.15.rst

@@ -860,6 +860,10 @@ New features
 
   (Contributed by Victor Stinner in :gh:`129813`.)
 
+* Add :c:func:`PyDict_NewPresized` function to create a dictionary with
+  preallocated items.
+  (Contributed by Victor Stinner in :gh:`139772`.)
+
 * Add :c:func:`PyTuple_FromArray` to create a :class:`tuple` from an array.
   (Contributed by Victor Stinner in :gh:`111489`.)
 

Include/cpython/dictobject.h

@@ -64,7 +64,13 @@ static inline Py_ssize_t PyDict_GET_SIZE(PyObject *op) {
 
 PyAPI_FUNC(int) PyDict_ContainsString(PyObject *mp, const char *key);
 
-PyAPI_FUNC(PyObject *) _PyDict_NewPresized(Py_ssize_t minused);
+PyAPI_FUNC(PyObject*) PyDict_NewPresized(Py_ssize_t size, int unicode_keys);
+
+Py_DEPRECATED(3.15) static inline PyObject*
+_PyDict_NewPresized(Py_ssize_t size)
+{
+    return PyDict_NewPresized(size, 0);
+}
 
 PyAPI_FUNC(int) PyDict_Pop(PyObject *dict, PyObject *key, PyObject **result);
 PyAPI_FUNC(int) PyDict_PopString(PyObject *dict, const char *key, PyObject **result);

Lib/test/test_capi/test_dict.py

@@ -545,6 +545,37 @@ def test_dict_popstring(self):
         # CRASHES dict_popstring({}, NULL)
         # CRASHES dict_popstring({"a": 1}, NULL)
 
+    def test_dict_newpresized(self):
+        # Test PyDict_NewPresized()
+        dict_newpresized = _testcapi.dict_newpresized
+
+        # non-unicode keys
+        d = dict_newpresized(3, 0)
+        d[1] = 'a'
+        d[2] = 'b'
+        d[3] = 'c'
+        self.assertEqual(d, {1: 'a', 2: 'b', 3: 'c'})
+
+        # unicode keys
+        d = dict_newpresized(3, 1)
+        d['a'] = 1
+        d['b'] = 2
+        d['c'] = 3
+        self.assertEqual(d, {'a': 1, 'b': 2, 'c': 3})
+
+        # mis-use unicode_keys parameter
+        d = dict_newpresized(3, 1)
+        d[1] = 'a'
+        d[2] = 'b'
+        d[3] = 'c'
+        self.assertEqual(d[2], 'b')
+
+        # "large" dictionary (1024 items)
+        d = dict_newpresized(3, 1)
+        for i in range(1024):
+            d[str(i)] = i
+        self.assertEqual(d, {str(i):i for i in range(1024)})
+
 
 if __name__ == "__main__":
     unittest.main()

Misc/NEWS.d/next/C_API/2025-10-08-15-26-21.gh-issue-139772.HBZeQA.rst

@@ -0,0 +1,2 @@
+Add :c:func:`PyDict_NewPresized` function to create a dictionary with
+preallocated items. Patch by Victor Stinner.

Modules/_testcapi/dict.c

@@ -258,6 +258,18 @@ test_dict_iteration(PyObject* self, PyObject *Py_UNUSED(ignored))
 }
 
 
+static PyObject *
+dict_newpresized(PyObject *self, PyObject *args)
+{
+    Py_ssize_t size;
+    int unicode_keys;
+    if (!PyArg_ParseTuple(args, "ni", &size, &unicode_keys)) {
+        return NULL;
+    }
+    return PyDict_NewPresized(size, unicode_keys);
+}
+
+
 static PyMethodDef test_methods[] = {
     {"dict_containsstring", dict_containsstring, METH_VARARGS},
     {"dict_getitemref", dict_getitemref, METH_VARARGS},
@@ -268,7 +280,8 @@ static PyMethodDef test_methods[] = {
     {"dict_pop_null", dict_pop_null, METH_VARARGS},
     {"dict_popstring", dict_popstring, METH_VARARGS},
     {"dict_popstring_null", dict_popstring_null, METH_VARARGS},
-    {"test_dict_iteration",     test_dict_iteration,             METH_NOARGS},
+    {"test_dict_iteration", test_dict_iteration, METH_NOARGS},
+    {"dict_newpresized", dict_newpresized, METH_VARARGS},
     {NULL},
 };
 

Objects/dictobject.c

@@ -2191,8 +2191,8 @@ dictresize(PyDictObject *mp,
     return 0;
 }
 
-static PyObject *
-dict_new_presized(Py_ssize_t minused, bool unicode)
+PyObject *
+PyDict_NewPresized(Py_ssize_t minused, int unicode_keys)
 {
     const uint8_t log2_max_presize = 17;
     const Py_ssize_t max_presize = ((Py_ssize_t)1) << log2_max_presize;
@@ -2213,17 +2213,12 @@ dict_new_presized(Py_ssize_t minused, bool unicode)
         log2_newsize = estimate_log2_keysize(minused);
     }
 
-    new_keys = new_keys_object(log2_newsize, unicode);
+    new_keys = new_keys_object(log2_newsize, unicode_keys);
     if (new_keys == NULL)
         return NULL;
     return new_dict(new_keys, NULL, 0, 0);
 }
 
-PyObject *
-_PyDict_NewPresized(Py_ssize_t minused)
-{
-    return dict_new_presized(minused, false);
-}
 
 PyObject *
 _PyDict_FromItems(PyObject *const *keys, Py_ssize_t keys_offset,
@@ -2241,7 +2236,7 @@ _PyDict_FromItems(PyObject *const *keys, Py_ssize_t keys_offset,
         ks += keys_offset;
     }
 
-    PyObject *dict = dict_new_presized(length, unicode);
+    PyObject *dict = PyDict_NewPresized(length, unicode);
     if (dict == NULL) {
         return NULL;
     }